Research And Development Of Intelligent Diagnosis System For Sucker Rod Pumping Units Based On Edge Computing

Oil extraction of sucker rod pumping units(SRPUs)is one of the most important production stages when oil wells are commissioned and thereafter.Timely fault detection and diagnosis of sucker rod pumping units is of great significance to maintain plant safety and improve the production efficiency.An intelligent diagnosis system for SRPUs based on edge computing is proposed in this paper.Based on the actual production data from crude oil extraction enterprises in northeast China,fault detection and diagnosis methods for SRPUs combined with statistical analysis and deep learning are studied on the cloud-edge collaboration framework to promote the stability in oil production,and at the meantime push the intelligent manufacturing construction of the crude oil extraction industry in China.In the first chapter,the background,significance and the principle of oil extraction technology using SRPUs,as well as the research status of the SRPUs fault diagnosis based on dynamometer cards are summarized,and then the inherent challenges of this study are analyzed.The main structure and content arrangement of this paper are given in the end.In the second chapter,the overall design of the intelligent diagnosis system for SRPUs based on edge computing is described.The system design objectives and overall architecture plan are analyzed.According to the difference of oilfield equipment conditions,fault detection and diagnosis algorithms for SRPUs based on statistical analysis and deep learning are respectively studied.Finally,the overall scheme based on cloud-edge collaboration framework is designed.In the third chapter,the fault detection and diagnosis methods for SRPUs based on statistical analysis is studied.Firstly,the principle and implementation steps are given which uses principal component analysis to model the long-term steady state of SRPUs and detect the irregular state,and then a method of SRPUs diagnosis based on characteristics of dynamometer cards and "four-point method" is studied.The effectiveness on fault detection and diagnosis of methods proposed in this chapter is proven by a case at last.In the fourth chapter,the fault detection and diagnosis methods for SRPUs based on deep learning is studied.Firstly,lightweight and efficient network of deep learning is analyzed and applied to a similarity judgment method of superimposed dynamometer cards.Data preprocessing,model training and testing processes are then elaborated.The fault detection and retrieval method based on triple loss is also studied,and the algorithm implementation process and experimental analysis are given in this chapter.Finally,the fault detection and diagnosis method for SRPUs based on similar recognition,which is a data fusion method of the two models,is proposed to improve the accuracy and real-time performance.In the fifth chapter,the intelligent diagnosis system for SRPUs based on edge computing is designed and developed.Firstly,the overall architecture and model deployment process of the edge computing gateway,as well as the development of cloud-edge communication are described.Then the database architecture of the cloud platform and modules such as real-time monitoring of oil wells,dynamometer cards setting and fault alarm management are developed.Finally,the actual operational effect is demonstrated.The sixth chapter is the summary and prospect of the whole paper.This chapter summarizes the research work,the existing problems and shortcomings,and the future research directions of this subject are pointed out in the end.The intelligent diagnosis system for SRPUs based on edge computing developed in this paper can effectively reduce the false alarm rate,solve the problems existing in the diagnosis methods based on pattern classification,and realize efficient and reliable intelligent diagnosis for SRPUs.This study has achieved the expected design objectives in the testing environment of crude oil extraction enterprises,and the system is being prepared for deployment and implementation,which surely has promising industrial application value in the future.